Sign Language Caption Generation Using LSTM

Automated sign language caption generation systems are vital since they help in communicating to people using sign language and others who are different from them. The work aims at proffering a novel approach for producing sign language translations using Tensorflow, Keras, and LSTM layers. To begin with, we used MediaPipe Holistic to identify important key points from sign-language video clips. Then we build a sign language model which is made up of LSTM layers on the basis of an improved action detection model that reads out the key points and predicts the sign language gestures in real time.